Originally published: SpaceNews
Vladimir Baranov is the Chief Information Officer and Chief Operating Officer at Scout. Scott Nelson is Scout’s Vice President of Product and Strategy.
This spring, more than 1,000 top tech leaders and expert researchers sounded the alarm on artificial intelligence (AI) and called for regulation on the emerging technology’s “profound risks to society and humanity,” in an open letter published by Future of Life Institute.
Should we also fear AI when it comes to our safety and security in space? Or should we be excited about the possibilities it might bring?
What is generative AI?
ChatGPT is part of OpenAI’s family of large language models, often abbreviated as LLMs. These newer breeds of AI models emerged around 2018 and fall under the category of “generative AI” — capable of learning patterns and structures in human language or image content and can have powerful reasoning and problem-solving capabilities. While the technology world has seen some impressive accomplishments in the world of AI over the past 2 decades, we will likely see more impact from these new models in the next 1-3 years than in the last 10 combined. Nearly every industry will see new and powerful AI systems try to reshape it.
However, most people are still looking at these new AI models as fun or convenient tools that can help with things such as recipes and writing articles. The truth: LLMs have ushered in a completely new paradigm in software engineering and will likely have a significant impact on the space industry, both by increasing the speed of development but also due to the vast cybersecurity risks these technologies will pose to our in-space infrastructure.
Risks for the space domain
The U.S. space industry is facing a rapidly evolving technology landscape. Many nations and adversaries will be drastically increasing their budgets and focus on developing AI capabilities, as the world is (slowly) coming to understand how much of a leap has been made even just in the last 5 years. We believe that U.S. space agencies (both civil and defense) and commercial organizations will need to be equally determined to better understand, develop, and deploy new capabilities to be able to defend themselves from potential threats.
Furthermore, the rise of LLMs will enable many, previously immature adversaries, to enhance their capabilities, posing new challenges to established powers. The rise of LLMs feels reminiscent of the early days of the internet, where our national security was attacked by individuals (i.e., Solar Sunrise–1998) or nation states who adopted and matured their technological capabilities quickly.
The most profound impact that generative AI will likely have is in code generation and system development. The combination of fine-tuning models for specific tasks, new architectures emerging on a weekly basis, and developers gaining experience with implementing models will lead to significant automation. As of today, boilerplate code and simple applications are being auto-written but we don’t see why entire production codebases couldn’t be autonomously generated by themselves in the near future. This would dramatically increase the productivity of teams and free up engineers to build at higher levels of abstraction. One likely outcome is that we could see more vertical integration coming from teams as the output of their software and hardware teams increases.
One of the core strengths of new AI models is their ability to create, rewrite, review, and summarize written artifacts. Given the enormous amount of contracting and other administrative efforts happening within space agencies, we can see a clear path toward drastically improving their ability to manage the vast amount of data they process, filter out noise, and equip their workforce with only the most important information in real time. This should have broader benefits to the commercial space industry, with faster and more efficient contract processing and capital allocation coming to mind.
Another benefit of extracting intelligence from vast amounts of data will be our ability to provide significantly better and actionable information to operators, allow them to manage more systems at once, and help them identify the best strategic decisions. Any data stream in today’s world can now be analyzed in seconds by highly capable algorithms to provide valuable strategic recommendations.
To effectively capture the emerging value of modern AI models, we see a couple first logical steps that most organizations should pursue.
* Educate and train
Invest in comprehensive education and training to ensure employees understand how these technologies work, their pros and cons, and their limitations. Allow teams to explore building with these systems in a safe and controlled environment to foster internal knowledge of their capabilities and how systems could be enhanced to improve our space capabilities. Also, foster collaboration and partnerships with industry and academia to increase awareness and knowledge transfer.
* Establish guidelines for ethical and responsible usage
As previously stated, there are numerous risks to developing AI technologies and no shortage of well-respected individuals who are urging caution at a global scale as technology companies continue to push the boundaries of what LLMs can do. It is therefore pertinent to combine education with strict guidelines that reflect ethical and responsible usage of such systems. Ideally, experts should be consulted to determine how to avoid their misuse.
As we find ourselves amidst a tectonic technology shift, adopting AI systems will undoubtedly prove critical in protecting our national security and our space infrastructure. By increasing our awareness and education of these technologies, we can aim to safely enhance our operational effectiveness and ensure our space agencies are equipped to operate and thrive in this rapidly changing environment. It is unclear what the exact impact of these technologies will be in the medium to long term, and how the space industry, in particular, will change as a result. However, we have seen how quickly progress is happening and wanted to share our understanding with a wider audience to foster discussions about how the space industry can best adopt AI safely and responsibly.
This article originally appeared in the May 2023 issue of SpaceNews magazine.